DEV Community

4663437Mehdi
4663437Mehdi

Posted on Originally published at 4663437mehdi.github.io

Token Ledger Digest – 2026-08-22

#ai #llm #api #news

Token Ledger Digest – 2026-08-22

Most cost‑impacting change: OpenAI’s GPT‑5.6 Sol Pro (and its non‑batch variant) dropped prompt pricing from $2.50 to $2.00 per 1M tokens and completion from $15.00 to $10.00 per 1M tokens – a $5.50/M token saving. Anyone running high‑volume completion workloads should see immediate cost cuts.

Added (9)

  • Meta: Muse Spark 1.2 Contributor – prompt $0.10/M, completion $0.20/M.
  • DeepSeek: DeepSeek V4 Flash Vision Exp – prompt $0.44/M, completion $1.32/M.
  • Ox Alpha – free (0/0).
  • Tencent: Hy‑MT2‑1.8B – prompt $0.044/M, completion $0.177/M.
  • Tencent: Hy‑MT2‑30B‑A3B – prompt $0.074/M, completion $0.295/M.
  • Thinking Machines: Inkling Small (free) – free.
  • Thinking Machines: Inkling (free) – free.
  • Mistral: Ministral 8B – prompt $0.11/M, completion $0.11/M.
  • Mancer: Weaver (alpha) – prompt $0.50/M, completion $0.75/M.

Removed (2)

  • Deep Cogito: Cogito v2.1 671B – prompt $1.25/M, completion $1.25/M (no longer available).
  • OpenAI: gpt‑oss‑20b (free) – free tier withdrawn.

Price Changes (18)

  • OpenAI: GPT‑5.6 Sol Pro – prompt ↓$2.50→$2.00/M, completion ↓$15.00→$10.00/M (‑$5.50/M).
  • OpenAI: GPT‑5.6 Sol – same ‑$5.50/M.
  • OpenAI GPT Latest – same ‑$5.50/M.
  • DeepSeek: DeepSeek V4 Pro 0423 – prompt ↓$1.60→$0.4475/M, completion ↓$3.20→$0.8951/M (‑$3.46/M).
  • MoonshotAI: Kimi K2.6 – prompt ↓$0.95→$0.5415/M, completion ↓$4.00→$2.28/M (‑$2.13/M).
  • OpenAI: GPT‑5.6 Sol Pro (batch) – prompt ↓$1.25→$1.00/M, completion ↓$7.50→$5.00/M (‑$2.75/M).
  • OpenAI: GPT‑5.6 Sol (batch) – same ‑$2.75/M.
  • MiniMax: MiniMax M2.5 – prompt ↑$0.225→$0.27/M, completion ↑$0.90→$1.08/M (+$0.225/M).
  • Google: Gemma 4 31B – prompt ↑$0.09→$0.10/M (+$0.01/M).
  • DeepSeek: DeepSeek V4 Flash 0731 – prompt ↓$0.14→$0.08/M, completion ↓$0.28→$0.18/M (‑$0.16/M).
  • MoonshotAI: Kimi K2.7 Code – prompt ↓$0.71→$0.67/M, completion ↓$3.50→$3.40/M (‑$0.14/M).
  • DeepSeek: DeepSeek V4 Flash – prompt ↓$0.0886→$0.0713/M, completion ↓$0.1772→$0.1425/M (‑$0.052/M).
  • DeepSeek: DeepSeek V3.2 – prompt ↓$0.269→$0.26/M, completion ↓$0.40→$0.38/M (‑$0.029/M).
  • DeepSeek V4 Flash Latest – completion ↑$0.14→$0.18/M (+$0.04/M).
  • All other changes are under $0.05/M and mainly affect niche tiers.

Who should care

  • Cost‑focused teams – prioritize the OpenAI Sol Pro/Sol drops and DeepSeek V4 Pro cuts.
  • Vision‑heavy users – new DeepSeek V4 Flash Vision Exp adds multimodal capability at moderate cost.
  • Free‑tier seekers – Ox Alpha, both Inkling models remain zero‑cost.
  • Batch workloads – OpenAI Sol Pro/Sol batch cuts halve expenses.
  • Budget‑conscious projects – the cheapest models today are inclusionAI Ling‑2.6‑flash ($0.01/$0.03/M), IBM Granite 4.0 Micro ($0.017/$0.112/M), and Mistral Nemo ($0.019/$0.03/M).

Total models tracked: 421.

Originally published at The Token Ledger. Subscribe for the daily digest.

Top comments (0)